Perceived Fairness of AI-Based Selection Tools

Abstract

This study examined job candidates’ perceived fairness of artificial intelligence–based selection tools within hiring contexts. Using a 3 × 2 between-subjects experimental design, 194 U.S. job seekers recruited through Prolific were randomly assigned to one of six vignette conditions varying by decision-maker agent (Human, Algorithmic, or Combined Human-Algorithm) and selection phase (Initial or Final). Perceived fairness was measured with the Selection Procedural Justice Scale (SPJS; Bauer et al., 2001), and general attitudes toward AI were assessed using the General Attitudes Toward Artificial Intelligence Scale (GAAIS; Schepman & Rodway, 2020). A factorial ANOVA revealed a significant main effect for decision-maker agent, F(2, 191) = 6.47, p = .002, with human decision-makers rated as significantly fairer than either algorithmic or combined agents. No significant effects were found for selection phase or for the interaction between agent type and phase. Moderation analyses using Hayes’ PROCESS macro indicated that general attitudes toward AI did not significantly moderate these relationships. Results suggest selection methods that utilize human decision-makers are perceived fairer than those that use AI, either solely or as an assistive decision-maker with human oversight. Findings underscore the need for transparency and meaningful human discretion in AI-supported hiring to maintain perceptions of procedural justice.